Editorial policy
Our standards for research, testing, fact-checking, AI use, and corrections.
Research
Every guide begins with primary sources: the official documentation of the project (n8n docs, Docker docs, WireGuard man pages, the project's GitHub README and changelog). Secondary sources are vendor blog posts, release notes, and well-known practitioner posts. Forum threads are used only as starting points, never as the final authority.
Testing
No tutorial is published before it has been executed end-to-end on a clean, minimal Ubuntu 22.04 or 24.04 VPS. We log every command, capture the final state, and verify the user-visible outcome (working URL, valid certificate, healthy container, expected log output). Tutorials that touch security are tested both with and without the security control to verify it actually does what we claim.
AI-assisted content policy
We use large language models as research and drafting assistants — never as a replacement for testing or for a human author. Specifically:
- An LLM may help outline a guide or draft a first pass.
- A human author rewrites the draft, runs every command on a live server, and rewrites again to match the actual observed behavior.
- No guide is published if the final article was not re-tested by a human on the latest stable version of the underlying software.
- Code blocks, version numbers, and command flags are never trusted from an LLM — they are taken from a real shell session.
This policy is aligned with Google's Helpful Content guidance: the value to the reader, not the production method, is what matters, and value requires verification.
Fact-checking
Numbers (RAM usage, bandwidth, RTT, prices, version numbers) are taken from the test server and labeled with the date measured. We do not cite "industry averages" without a primary source. Vendor claims (e.g., "3.5 Tbps DDoS protection") are attributed to the vendor and not presented as our own measurement.
Corrections
Errors are corrected in-place with a dated changelog entry at the bottom of the affected article. Material corrections (changed recommendation, security-impacting fix) are flagged at the top of the article.
Update cadence
Every guide is reviewed at least once every six months and re-tested when the underlying project ships a major version. The "last tested" date is visible on every guide.
Conflict of interest
See our affiliate disclosure for the full picture. In short: we recommend hosting providers we have actually used in production. We disclose every affiliate relationship.